- 16th April 2018
- Posted by: Manolis
Not only Google and Facebook should have access to AI, it argues
BOSSES are more likely to groan than feel giddy about advances in artificial intelligence (AI). They need a strategy, but few companies can hope to own a unit like Google’s DeepMind, whose algorithms not only beat the world’s best Go players but made a 40% improvement in the energy efficiency of its parent’s data centres. A Canadian startup, Element AI, wants to let all businesses tap into the world’s best AI minds.
The brain behind the new firm is Yoshua Bengio, a pioneer in “deep learning”, a branch of AI. As firms such as Google and Facebook lured dozens of AI academics, some in the field expressed fears about a brain drain from academia. In 2015, for example, Uber, a ride-hailing startup, poached 40 researchers from Carnegie Mellon University. Mr Bengio meanwhile stayed at the University of Montreal (though in January he became an adviser to Microsoft)
Element AI will let researchers stay in their university posts while working on corporate projects. It plans, in effect, to build an AI platform on which a network of member firms (in which it may take stakes) can serve other companies. These member firms will tap Element AI’s brain trust and license its technical platform. This month the startup raised $102m of capital from backers including Intel and Nvidia, two chip giants.
Its system addresses a shortcoming of many AI applications. Individual firms are awash with data but may not have enough to train AI models. Element AI’s network will be able to share algorithmic learning from all the data they crunch, enabling better performance than they would achieve using only one client’s data. For example, an oil major might want to use image-recognition to identify corrosion on its pipes. Element AI could develop a system to spot it and predict the likelihood of a leak, to rank which pipes get fixed first. If the client lacks images to train the algorithm, Element AI’s work in an adjacent area—say, corrosion on railway tracks—could be used.
Jean-François Gagné, Element AI’s boss, says that the company aims to “democratise” AI by making state-of-the-art technology available to companies well beyond the main technology giants. “We are a neutral player you can trust,” he argues. But it is notoriously hard to move techniques from the research lab into real-life applications.
If AI does become the bedrock of corporate technology, there should be room for several models. Big consultancies are already believers and have begun acquiring data-analytics firms themselves. Element AI’s approach is promising. But the McKinsey of AI may yet turn out to be McKinsey itself.